Artificial intelligence (AI) for reaction condition optimization has become an important topic in the pharmaceutical industry, given that a data-driven AI model can assist drug discovery and accelerate reaction design. However, existing AI models lack the chemical insights and real-time knowledge acquisition abilities of experienced human chemists. This paper proposes a Large Language Model (LLM) empowered AI agent to bridge this gap. We put forth a novel three-phase paradigm and applied advanced intelligence-enhancement methods like in-context learning and multi-LLM debate so that the AI agent can borrow human insight and update its knowledge by searching the latest chemical literature. Additionally, we introduce a novel Coarse-label Contrastive Learning (CCL) based chemical fingerprint that greatly enhances the agent's performance in optimizing the reaction condition. With the above efforts, the proposed AI agent can autonomously generate the optimal reaction condition recommendation without any human interaction. Further, the agent is highly professional in terms of chemical reactions. It demonstrates close-to-human performance and strong generalization capability in both dry-lab and wet-lab experiments. As the first attempt in the chemical AI agent, this work goes a step further in the field of "AI for chemistry" and opens up new possibilities for computer-aided synthesis planning.